Computer-implemented method for estimating road surface properties

Abstract

Computer-implemented method for estimating road surface properties of a roadway to be traveled on by a motor vehicle (8), comprising the following steps: - Collection of meteorological data (6) along the roadway; - Estimation of road surface properties using artificial intelligence with meteorological data (6), characterized in that the meteorological data (6) comprise images and / or videos recorded by cameras (5) located remotely from the motor vehicle (8), wherein, during travel on the road surface, sensors of the motor vehicle (8) record measured values ​​that provide an indication of the road surface properties, wherein the measured values ​​and the road surface properties estimated by the artificial intelligence are transmitted to a computer (9) located remotely from the motor vehicle (8), wherein the computer (9) determines and provides the road surface properties of the different roads from the estimated road surface properties of several different roads and the measured values ​​of several motor vehicles.

Description

[0001] The present invention relates to a computer-implemented method for estimating road surface properties according to the preamble of claim 1.

[0002] Such methods are known from the prior art. US patent 2019 / 0217864 A1 discloses a method for estimating road surface properties along a route using artificial intelligence.

[0003] DE 10 2018 203 807 A1 relates to a method for classifying a road surface condition based on image data from a vehicle camera system and to a corresponding vehicle camera system. Further methods are known from DE 10 2015 100 812 A1 and DE 10 2013 226 631 A1.

[0004] In contrast, the present invention aims to enable a more accurate assessment of road surface properties.

[0005] This problem is solved by a method according to claim 1. Embodiments of the invention are specified in the dependent claims.

[0006] The method according to claim 1 serves to estimate the road surface characteristics of a roadway to be traveled on by a motor vehicle. This could, for example, be a roadway that the motor vehicle will be traveling on in the immediate future. However, it is also possible that it is a roadway located far away from the motor vehicle, which is included in a route plan on the way to a destination.

[0007] The procedure involves collecting meteorological data along the roadway. For the purposes of this description, meteorological data refers specifically to data relating to the weather at the location of the roadway to be traversed. This can include, for example, temperature, air pressure, precipitation intensity, type of precipitation, wind speed, wind direction, and / or information about cloud cover.

[0008] Using meteorological maps, an artificial intelligence then estimates the road surface properties. Since the meteorological data used by the AI ​​only indirectly influences the road surface properties, this is an estimate. However, this estimate can be particularly accurate if, for example, information about the road surface is also used.

[0009] The assessment can also categorize road surface characteristics. For example, it can be divided into categories such as "high risk" and "low risk." Road surfaces categorized as "high risk" can then potentially be avoided when driving a motor vehicle.

[0010] According to the invention, the meteorological data includes images and / or videos recorded by cameras mounted remotely from the motor vehicle. For the purposes of this description, this refers in particular to stationary cameras that do not move relative to their surroundings during operation. These cameras could, for example, be cameras used to monitor traffic on the roadway. Such cameras are already present in many countries, so it is not necessary to install additional cameras.

[0011] Meteorological data can be extracted from the images and / or videos by analyzing whether it is raining, snowing, or hailing. The intensity of the respective precipitation can also be analyzed. Furthermore, it is possible to estimate how much snow or water is present on the road. Using the images and / or videos is advantageous for increasing the accuracy of the estimation without requiring additional cameras that are not currently available. Of course, installing new cameras for this purpose is also possible.

[0012] The assessment of road surface characteristics can preferably be used by other components of the vehicle. This is particularly advantageous for driver assistance systems and / or a system that controls the vehicle fully automatically. In the context of this description, this specifically means that no user interaction is necessary to safely control the vehicle. Such a system can, for example, signal to a driver that they must take control of the vehicle when the vehicle approaches a road surface classified as "high risk".

[0013] While driving on the road, the vehicle's sensors collect data that provides information about the road surface properties. These measurements, along with the road surface properties estimated by the artificial intelligence, are transmitted to a computer located remotely from the vehicle. This is particularly advantageous because the computer can then further utilize this data. For example, the computer can use the measurements to verify the accuracy of the estimate. If the measurements indicate a lack of accuracy, the estimate can be reconsidered. It is also possible for the computer to discard the AI's estimate as incorrect, reassess the road surface properties based on the measurements, and transmit the result of this reassessment back to the vehicle.

[0014] The computer uses estimated road surface properties from several different road surfaces, particularly different roads, and measurement data from multiple vehicles to determine and make available the road surface properties of these different surfaces. This embodiment could, for example, be a cloud-based method. Through the transmitted measurement data and the estimated road surface properties, the computer can make the determined road surface properties available to a large number of vehicles, which can then, for example, adjust a planned route to avoid roads with undesirable road surface properties.

[0015] According to one embodiment of the invention, the cameras can be configured as real-time cameras. For the purposes of this description, this means in particular that the cameras record images and / or videos and send them immediately after recording to a server located remotely from the camera. The use of such cameras is particularly advantageous for increasing the quality of the assessment, since the images and / or videos are especially up-to-date.

[0016] According to one embodiment of the invention, the meteorological data can include publicly available data from a weather service. For the purposes of this description, publicly available data is understood to mean, in particular, data that can be used by anyone free of charge or for a fee.

[0017] It is possible that the assessment is carried out using map data. For example, the map data can assign a road class, an elevation above sea level, and / or a number of lanes to the roadway. The road class can, for example, define the type of road to which the roadway belongs. The road class can be selected from the following categories, for example: motorway, federal highway, rural road, county road, rural road.

[0018] The map data can be used, for example, to make the estimation more accurate. Furthermore, the map data can be used, for example, to define a geographical area of ​​validity for the use of the meteorological data, particularly the images and / or videos. For instance, the images and / or videos can be used for a relatively large area of ​​validity if the road is classified as a "motorway," whereas they will have a smaller area of ​​validity if the road is classified as a "federal highway" or "state road." The area of ​​validity could, for example, correspond to the length of the roadway.

[0019] Furthermore, meteorological data can include vehicle data. This vehicle data can be generated, for example, by vehicle components. Specifically, it can include data from an outside temperature sensor, data regarding windshield wiper activity, images and / or video from a vehicle camera, data from an ultrasonic sensor, acoustic data, data from a wheel speed sensor, and / or data from a vehicle dynamics control system such as ABS and / or ESP. All of this data allows for a more accurate assessment of road surface characteristics.

[0020] According to one embodiment of the invention, the artificial intelligence can be self-learning. The AI ​​can be trained with training data before estimating the road surface properties. This training data can include meteorological training data and surface data of a training road surface. In this way, the AI ​​can learn how specific meteorological training data influences the surface data. The surface data could, for example, be friction coefficients between the training road surface and the tires of a training vehicle. The surface data can be acquired visually. For example, it can be specified that the road is slightly or heavily wet, icy, snow-covered, or dry. The training road surface could, for example, be a test track or a regular road.

[0021] According to one embodiment of the invention, surface data can be acquired using a sensor on a training vehicle. The sensor can be, for example, a camera and / or an ultrasonic sensor. Surface data acquired by the sensor is advantageously objective and less subjective than surface data acquired visually – for example, by a user.

[0022] According to one embodiment of the invention, the estimated road surface properties can be used to plan a route from a starting point to a destination point. The route can be planned, for example, for a fully automated vehicle or for a vehicle with a human driver.

[0023] According to one embodiment of the invention, the estimation of road surface properties can be performed by a computer located remotely from the motor vehicle. The computer can also transmit the estimated road surface properties to the motor vehicle. In particular, this can be the same computer that determines and provides the road surface properties. However, it is also possible for the two computers to be different. Further features and advantages of the present invention will become clear with reference to the following description of preferred embodiments and the accompanying figures. The same reference numerals are used for identical or similar features and for features with identical or similar functions. Fig. 1 a schematic block diagram of part of a method according to an embodiment of the invention; Fig. 2 a schematic block diagram of part of a method according to an embodiment of the invention; and Fig. 3 a schematic block diagram of part of a method according to an embodiment of the invention.

[0024] In Fig. Figure 1 illustrates how meteorological training data 1 and surface data of a training track are initially acquired by a sensor on a training vehicle 2. The meteorological training data 1 can, for example, be data from a publicly accessible database. Alternatively or additionally, the meteorological data can be acquired by sensors located near the training track. The meteorological data can include, for example, precipitation amount, precipitation type, wind speed, cloud cover, and / or temperature. It is particularly possible that the meteorological data may also include historical data. For example, the meteorological data can cover a period of one hour. The meteorological data thus allows conclusions to be drawn about the road surface properties of the training track.In particular, it is possible to assess whether the training track is wet, dry, covered with snow or icy.

[0025] The sensor of the training vehicle can be, for example, an ultrasonic sensor or a camera. The surface data recorded by the sensor of training vehicle 2 allows conclusions to be drawn about the road surface properties of the training track. In particular, it can be estimated whether the training track is wet, dry, covered with snow, or icy.

[0026] As an alternative to recording surface data with the sensor, the surface data can also be recorded by a person in training vehicle 2. The person can, for example, determine whether the training track is wet, dry, covered with snow, or icy.

[0027] The meteorological training data 1 and the surface data are used to train an artificial intelligence 3. The artificial intelligence can be trained to learn independently for this purpose. For example, the artificial intelligence can create a model 4 that estimates the road surface properties when meteorological data for a specific roadway is input.

[0028] Various methods can be used to create Model 4. For example, an agent can be used that is rewarded or punished depending on whether it correctly assesses the road surface characteristics of the training track. The agent can, for instance, categorize the road surface characteristics into different risk categories. The agent can be penalized if it chooses the wrong risk category. Depending on the intended use of Model 4, the punishment can be such that, for example, the agent is severely penalized for choosing too low a risk category and receives little or no punishment for choosing too high a risk category. This is advantageous if Model 4 is later to be used for a self-driving car. In this application, an underestimation of risk would be disastrous.

[0029] However, if Model 4 is intended to be used, for example, as an aid for a human driver of the vehicle, an incorrect risk assessment is detrimental but not fatal. Thus, in such a later application, an incorrect risk classification can be penalized in any case. However, the penalty need not necessarily be as severe as in the case mentioned above.

[0030] In Fig. Figure 2 illustrates the use of Model 4 during the operation of a motor vehicle. Meteorological data 6 are used as input for Model 4. The meteorological data 6 comprise images and / or videos recorded by a camera 5 located remotely from the motor vehicle. The camera 5 is positioned near the roadway traveled by the motor vehicle. The camera 5 could, for example, be a traffic camera that records images and / or videos in real time and uploads them to a server via a data connection.

[0031] The meteorological data 6 are used as input for the model 4. The road surface properties are estimated from this data. These estimated road surface properties can then be output to a technical device 7. The technical device 7 could, for example, be a digital processing unit that uses the estimated road surface properties, for instance, in planning a road for a motor vehicle. For example, the processing unit could be configured to plan a route from a starting point to a destination point. During this planning process, the road surface properties of different lanes can also be taken into account. For example, the road surface properties can classify each lane into a category. This category could, for instance, relate to the risk of an uncontrollable driving situation on the respective lane.This allows the processing unit, for example, to avoid high-risk lanes when planning the road to be traveled by the motor vehicle.

[0032] It is also possible for the road surface characteristics to be displayed to the vehicle user via a screen. This allows the user to see and, if necessary, select which road surface with which characteristics they are driving on.

[0033] The use of images and / or videos recorded by camera 5 is advantageous for a particularly high accuracy in estimating road surface properties.

[0034] It is particularly advantageous to use meteorological data for multiple lanes, thus enabling the estimation of road surface characteristics for several lanes. This allows the technical device to present a particularly comprehensive picture to the user and / or to consider a large number of road surface characteristics from different lanes when calculating the route from the starting point to the destination.

[0035] The in Fig. The method described in section 3 is an alternative to the method from Fig. 2. The meteorological data 6 and the images and / or videos from camera 5 are - as shown Fig. 2 known – by artificial intelligence using model 4 to estimate the road surface properties. Subsequently, measured values ​​acquired by sensors of the vehicle 8 and the estimated road surface properties are transmitted to a computer 9 located remotely from the vehicle 8. The measured values ​​can also include information about the road surface properties. For example, the measured values ​​can include information about the coefficient of friction between the vehicle's tires and the road surface. The measured values ​​can include speed, deceleration, acceleration, data from an anti-lock braking system, and / or data from an anti-slip control system. The computer 9 can, for example, be a server to which a large number of vehicles transmit their acquired measured values.

[0036] Computer 9 determines the road surface properties of each lane from the estimated road surface properties and the measured values. These road surface properties determined by Computer 9 can then be used – similarly to in Fig. 2 the road surface properties estimated by the artificial intelligence - are transferred to the technical device 7.

[0037] The technical device 7 can process the determined road surface properties in a similar way to the estimated road surface properties in Fig. 2.

Claims

[1] Computer-implemented method for estimating road surface properties of a roadway to be traveled on by a motor vehicle (8), comprising the following steps: - Collection of meteorological data (6) along the roadway; - Estimation of road surface properties using artificial intelligence and meteorological data (6), characterized by, that the meteorological data (6) comprise images and / or videos recorded by cameras (5) located remotely from the motor vehicle (8), wherein, while driving on the roadway, sensors of the motor vehicle (8) record measurements that provide an indication of the roadway properties, wherein the measurements and the roadway properties estimated by the artificial intelligence are transmitted to a computer (9) located remotely from the motor vehicle (8), wherein the computer (9) determines and provides the roadway properties of the different roadways from the estimated roadway properties of several different roadways and the measurements of several motor vehicles. [2] Method according to claim 1, characterized by , that the cameras (5) are designed as real-time cameras. [3] Method according to any one of the preceding claims, characterized by, that the meteorological data (6) include publicly available data from a weather service. [4] Method according to any one of the preceding claims, characterized by , that the artificial intelligence is trained in a self-learning manner, wherein the artificial intelligence is trained with training data (1) prior to the estimation of the road properties, and wherein the training data (1) comprise meteorological training data (1) and surface data of a training road. [5] Method according to the previous claim, characterized by , that the surface data are acquired using a sensor of a training vehicle (2). [6] Method according to any one of the preceding claims, characterized by , that the estimated road surface properties are used to plan a route from a starting point to a destination point. [7] Method according to any of the preceding claims, characterized by, that the estimation of the road surface properties is carried out by a computer located remotely from the motor vehicle, and wherein the computer transmits the estimated road surface properties to the motor vehicle (8).

Images